Electrochemical Immunochip Sensor for Aflatoxin M1 Detection

To whom correspondence should be addressed. Ibtisam E. Tothill: fax +44(0)1234 75 8380, e-mail [email protected]. Damien W. M. Arrigan: fax +3...
6 downloads 4 Views 2MB Size
Anal. Chem. 2009, 81, 5291–5298

Electrochemical Immunochip Sensor for Aflatoxin M1 Detection Charlie O. Parker,† Yvonne H. Lanyon,‡ Mary Manning,‡ Damien W. M. Arrigan,*,‡ and Ibtisam E. Tothill*,† Cranfield Health, Cranfield University, Cranfield, Bedfordshire, MK43 0AL, U.K., and Tyndall National Institute, Lee Maltings, University College Cork, Cork, Ireland An investigation into the fabrication, electrochemical characterization, and development of a microelectrode array (MEA) immunosensor for aflatoxin M1 is presented in this paper. Gold MEAs (consisting of 35 microsquare electrodes with 20 µm × 20 µm dimensions and edge-to-edge spacing of 200 µm) together with onchip reference and counter electrodes were fabricated using standard photolithographic methods. The MEAs were then characterized by cyclic voltammetry, and the behavior of the on-chip electrodes were evaluated. The microarray sensors were assessed for their applicability to the development of an immunosensor for the analysis of aflatoxin M1 directly in milk samples. Following the sensor surface silanization, antibodies were immobilized by cross-linking with 1,4-phenylene diisothiocyanate (PDITC). Surface characterization was conducted by electrochemistry, fluorescence microscopy, scanning electron microscopy (SEM), and atomic force microscopy (AFM). A competitive enzyme linked immunosorbent assay (ELISA) assay format was developed on the microarray electrode surface using the 3,3,5′,5′-tetramethylbenzidine dihyrochloride (TMB)/ H2O2 electrochemical detection scheme with horseradish peroxidase (HRP) as the enzyme label. The performance of the assay and the microarray sensor were characterized in pure buffer conditions before applying to the milk samples. With the use of this approach, the detection limit for aflatoxin M1 in milk was estimated to be 8 ng L-1, with a dynamic detection range of 10-100 ng L-1, which meets present legislative limits of 50 ng L-1. The milk interference with the sensor surface was also found to be minimal. These devices show high potential for development of a range of new applications which have previously only been detected using elaborate instrumentation. Electrochemical sensors are renowned for their excellent sensitivity, selectivity, versatility, and simplicity, and therefore there is a continual interest in their development for the analysis * To whom correspondence should be addressed. Ibtisam E. Tothill: fax +44(0)1234 75 8380, e-mail [email protected]. Damien W. M. Arrigan: fax +353-21-4270271, e-mail [email protected]. † Cranfield University. ‡ Tyndall National Institute. 10.1021/ac900511e CCC: $40.75  2009 American Chemical Society Published on Web 06/02/2009

of environmental, food, and clinical samples.1-6 Different types of sensor platforms have been used for electrochemical sensors, but most are based on screen-printed technology.7 Interest in the use of microelectrodes based on photolithographic techniques coupled with electrochemical detection methods is increasing.8-10 A microelectrode is described as an electrode where one of its dimensions is in the micrometer range.11 These developments and advances in sensor technology have been fuelled by medical applications where microelectrodes can be implanted to monitor electrophysiological pulses such as in cardiac tissues and also in other applications.12 One of the main benefits of using a microelectrode in a sensor application is the greater sensitivity that arises from the enhanced mass-transport at these small electrodes.9 Hemispherical diffusion layers are formed at such electrodes and a much faster diffusion of electroactive substances occurs due to the multidimensional nature of this process, resulting in sigmoidal (or steady-state) cyclic voltammograms (CVs).13,14 The advantages are in the improved response time (faster response), greater sensitivity and increased response per unit electrode surface area (greater current density, increasing the signal-to-noise ratio). However, this results in very low current values which can be problematic.11,14 The use of an array of microelectrodes addresses this problem by providing a substantial improvement in the signal-to-noise ratio under steady-state conditions.15,16 The spacing between the electrodes in these arrays (1) Tothill, I. E., Piletsky, S., Magan, N., Turner, A. P. F. In Instrumentation and Sensors for the Food Industry, 2nd ed.;Woodhead Publishing Limited CRC Press: Boca Raton, FL, 2001; pp 760-775. (2) Mascini, M. Pure Appl. Chem. 2001, 73, 23–30. (3) Wang, J. Acc. Chem. Res. 2002, 811–816. (4) Tothill, I. E., Turner, A. P. F. In Encyclopedia of Food Sciences and Nutrition, 2nd ed.; Academic Press: New York, 2003; pp 489-499. (5) Pemberton, R. M.; Mottram, T. T.; Hart, J. P. J. Biochem. Biophys. Methods 2005, 63, 201–212. (6) Bakker, E.; Qin, Y. Anal. Chem. 2006, 78, 3965–3983. (7) Newman, J. D.; Turner, A. P. F. Biosens. Bioelectron. 2005, 20, 2435–2453. (8) Berduque, A.; Lanyon, Y. H.; Beni, V.; Herzog, G.; Watson, Y. E.; Rodgers, K.; Stam, F.; Alderman, J.; Arrigan, D. W. M. Talanta 2007, 71, 1022– 1030. (9) Ordeig, O.; del Campo, J.; Mun ˜oz, F. X.; Banks, C. E.; Compton, R. G. Electroanalysis 2007, 19, 1973–1986. (10) Beni, V.; Arrigan, D. W. M. Current Anal. Chem. 2008, 4, 229–241. (11) Stulik, K.; Amatore, C.; Holub, K.; Marecek, V.; Kutner, W. Pure Appl. Chem. 2000, 72, 1483–1492. (12) Hoffman, B. F. Cardiovasc. Res. 2002, 53, 1–5. (13) Amatore, C. In Physical Electrochemistry: Principles, Methods and Applications; Rubinstein, I., Ed.; Marcel Dekker: New York, 1995; p 131. (14) Alden, J. A.; Booth, J.; Compton, R. G.; Dryfe, R. A. W.; Sanders, G. H. W. J. Electroanal. Chem. 1995, 389, 45–54. (15) Feeney, R.; Kounaves, S. P. Electroanalysis 2000, 12, 677–684.

Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

5291

is important and needs to be such that each element of the array experiences individual, noninteracting diffusion profiles.16-19 Then steady-state behavior can be achieved, which is best for sensor applications. The enhanced capability of microelectrode arrays (MEAs) as sensing devices makes them an ideal choice for trace analysis, such as aflatoxin M1 (AFM1) analysis. Aflatoxin M1 (Cyclopenta (C) furo(3′,2′:4,5)furo(2,3-H)(1)benzopyran-1,11-dione,2,3,6A,9A tetrahydro-9°-hydroxy-4-methoxy, CAS number 6795-23-9, chemical formula C17H12O7, relative molecular mass 328.3 amu) is excreted in milk by animals upon the digestion of feed contaminated with the fungal toxin aflatoxin B1.20,21 It has been theorized that aflatoxin M1 is a detoxification product of aflatoxin B1 since the carcinogenicity of aflatoxin M1 is lower than aflatoxin B1.22 However, aflatoxin M1 is still regarded as a carcinogenic, genotoxic, teratogenic, and immunosuppressive compound. Aflatoxin M1 can also be found in other dairy products such as cheese, yogurt, and infant formulas23 and also in human breast milk.24 Because of the fact that the milk intake in infants is high and that they are very vulnerable to toxins, the European Commission regulation 472/2002 imposes maximum permissible levels of aflatoxin M1 in milk of 50 ng L-1 and in infant formulas of 25 ng L-1.25 Determination of aflatoxin M1 is usually conducted using HPLC, TLC, and ELISA methods which are all laboratory-based systems and require the expertise of trained personnel.26-28 Unfortunately the regions of the world which are most affected by aflatoxin contamination tend to be poorer areas within the tropics. Therefore, as stipulated by the United Nations “there is an urgent need for simple, robust, low-cost analysis methods, for the major mycotoxins, which can be used in developing country laboratories”.29 Microfabricated sensor systems offer many benefits to achieve those goals.30 In this article, the development of an MEA-based immunosensor for aflatoxin M1 is reported. The chip-based electrochemical (16) Bard, A. J., Faulkner, L. R. In Fundamentals and Applications, 2nd ed.; John Wiley and Sons: New York, 2001; p 168. (17) Sandison, M.; Anicet, N.; Glidle, A.; Cooper, J. M. Anal. Chem. 2002, 74, 5717–5725. (18) Davies, T. J.; Compton, R. G. J. Electroanal. Chem. 2005, 585, 63–82. (19) Davies, T. J.; Ward-Jones, S.; Banks, C. E.; del Campo, J.; Mas, R.; Munoz, F. X.; Compton, R. G. J. Electroanal. Chem. 2005, 585, 51–62. (20) Sargeant, K.; Sheridan, A.; O’Kelly, J. Nature 1961, 192, 1096–1097. (21) Holzapfel, C. W.; Steyn, P. S. Tetrahedron Lett. 1966, 25, 2799–2803. (22) Neal, G. E.; Eaton, D. L.; Judah, D. J.; Verma, A. Toxicol. Appl. Pharmacol. 1998, 151, 152–158. (23) Martins, M. L.; Martins, H. M. Int. J. Food Microbiol. 2004, 91, 315–317. (24) El-Nezam, H. S.; Nicoletti, G.; Neal, G. E.; Donohue, D. C.; Ahokas, J. T. Food Chem. Toxicol. 1995, 33, 173–179. (25) Henry, S. H.; Whitaker, T.; Rabbani, I.; Bowers, J.; Park, D.; Price, W.; Bosch, F. X.; Pennington, J.; Verger, P.; Yoshizawa, T.; van Egmond, H.; Jonker, M. A.; Coker, R. Aflatoxin M1; Report 1012, (WHO Additives, Series 47), Joint Expert Committee on Food Additives (JECFA), 2001. (26) Kamkar, A. Food Control 2005, 16, 593–599. (27) Oveisi, M. R.; Jannat, B.; Sadeghi, N.; Hajimahmoodi, M.; Nikzad, A. Food Control 2007, 18, 1216–1218. (28) Rodriguez Velasco, M. L.; Calonge Delso, M. M.; Ordonez Escudero, D. Food Addit. Contam. 2003, 20, 276–280. (29) Proctor, D. L., Ed. Grain Storage Techniques: Evolution and Trends in Developing Countries; Food and Agriculture Organization of the United Nations: Rome, Italy, 1994. (30) Logrieco, A.; Arrigan, D. W. M.; Brengel-Pesce, K.; Siciliano, P.; Tothill, I. E. Food Addit. Contam. 2005, 22, 335–344. (31) Lucarelli, F.; Marrazza, G.; Turner, A. P. F.; Mascini, M. Biosens Bioelectron. 2004, 19, 515–530.

5292

Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

cell was fabricated to contain the working electrode, which was the MEA, and counter and reference electrodes, so that all necessary electrodes for electrochemical measurements were contained on the chip. The assay on the sensor chip was based on a competitive format between the free aflatoxin M1 in the sample and an aflatoxin-horseradish peroxidase conjugate for an immobilized monoclonal antibody for aflatoxin M1. With the use of chronoamperometry, the depletion of hydrogen peroxide was monitored via 3,3,5′,5′-tetramethylbenzidine dihyrochloride (TMB) mediation to ascertain the concentration of HRP on the sensor and consequently the concentration of aflatoxin M1 in the sample. EXPERIMENTAL SECTION Reagents and Solutions. Aflatoxin M1 was purchased from Axxora UK Limited (Nottingham, U.K.), anti-aflatoxin M1 antibody (raised from rat) was purchased from Abcam Limited, (Cambridge U.K.), and aflatoxin M1-HRP conjugate was obtained from a RIDASCREEN kit from R-Biopharm (Glasgow, U.K.). 3,3,5′,5′-Tetramethylbenzidine dihydrochloride, hydrogen peroxide, and Tween 20 were purchased from Sigma-Aldrich (Poole, U.K.). Anti-rat immunopure antibody (raised in goat with affinity for the Fc fragment only) was from Perbio Science (Cramlington, U.K.). Milk and dried milk samples were obtained from the local supermarket. All other chemicals were purchased from Sigma-Aldrich (Poole, U.K.) or otherwise as stated in the text. Microfabrication. Gold cell-on-a-chip microelectrodes (including on-chip reference and counter electrodes) were fabricated by standard deposition, etching, and lithographic techniques used in microfabrication technology. The first step involved growth of a thermal oxide on a silicon wafer. This was followed by plasma enhanced chemical vapor deposition (PECVD) of a silicon dioxide layer. Photoresist was then spun onto the wafer and patterned, and the exposed oxide layer was then wet etched. For the fabrication of the metal electrodes, gold was deposited by evaporation of Ti/Pt/Au multilayers in the proportion 30:50:250 nm and the remaining silicon dioxide was removed by a buffer oxide etch (HF and NH4F). This was followed by deposition of a Si3N4 passivation layer. The recessed microelectrode array (500 nm recess depth) was then obtained using a photolithographic etch process (Pt-EKC solvent). Following fabrication, the wafers were diced and the electrodes packaged on printed circuit boards (PCB) by attaching the individual chips to the PCB with silver epoxy die attach (Ablebond 8484, Ablestik), wire bonding the bondpads on the chips to the PCB with 25 µm aluminum wires, and finally protecting the wirebonds and chip edges by covering in a polymeric selective encapsulant (Amicon 50300 HT, Emerson & Cuming). The electrode device containing all necessary electrodes (working, counter, and reference) is referred to as the cell-on-chip device. The electrode materials were either gold or platinum, including the pseudoreference electrodes. All microfabrication processing was carried out at the Central Fabrication Facility at Tyndall National Institute (Cork, Ireland). (32) Yang, M.; Yau, H. C. M.; Chan, H. L. Langmuir 1998, 14, 6121–6129. (33) Ouerghi, O.; Touhami, A.; Othmane, A.; Ben Ouada, H.; Martelet, C.; Fretigny, C.; Jaffreezic Renault, N. Biomol. Eng. 2002, 19, 183–188. (34) Parker, C.; Tothill, I. E. Biosensors. Bioelectron. 2009, 24, 2452–2457.

Electrochemical Characterization and Reagents. A CHI620A electrochemical analyzer with picoamp booster and faraday cage (CH Instruments, Texas) was used for the electrochemical characterization studies. Experiments were performed either with external reference and counter electrodes with a Ag|AgCl reference electrode and platinum wire counter electrode (both from CH instruments) or using the on-chip (pseudo)reference and counter electrodes. Both on-chip pseudoreference and off-chip reference electrodes were used in the evaluation study so as to provide a realistic comparison to standard laboratory operations. Prior to electrochemical testing, the chips were treated with oxygen plasma for 10 min at 150 W to remove any residual organic matter. Cyclic voltammograms (CVs) were recorded at the Au MEAs in 1 mM ferrocene monocarboxylic acid (FcCOOH) in 0.01 M phosphate buffered saline (PBS) solution. Characterization of DNA-surface modified Au or Pt electrodes was performed by CV at 5 mV s-1 using 5 mM ferricyanide solution in 0.1 M potassium chloride. On-chip reference electrode preparations were undertaken by modification of the gold pseudoreference electrodes by electrodeposition of silver from an aqueous 5 mM silver nitrate solution in 50 mM potassium nitrate and 0.5 M potassium thiocyanate. Electrodeposition of silver was achieved at a fixed potential of -0.15 V (vs a Ag wire) for 10 min. Silver/silver chloride (Ag/AgCl) was formed by the immersion of the on-chip silver electrode in 1 M iron(III) chloride for 60 s. Surface Modification. Surface modification of the chips was performed as follows: (i) pretreatment of the chips with oxygen plasma (150 W, 10 min); (ii) silanization by immersion of the chips in 3% 3-aminopropyltrimethoxysilane (APTES) (Gelest) in a 19:1 dilution of methanol/deionized (DI) water, followed by washing with methanol and DI water; (iii) heat curing of the chips at 120 °C for 15 min; (iv) deposition of a cross-linker by immersion of the chips in dimethylformamide (DMF) containing 10% pyridine and 1 mM 1,4-phenylene diisothicyanate (PDITC) (Fluka) for 2 h; (v) final washing of the chips with DMF and 1,2 dichloroethane followed by drying under a stream of nitrogen. The primary immunoreagents of the sensor were covalently immobilized following this stage of the surface modification (Antibody Immobilization onto the Chip Device). Characterization and assessment of the surface functionalization was undertaken using fluorescently tagged ssDNA (single stranded DNA with an amine anchor on the 5′ end, for attachment to the PDITC cross-linker, and a fluorescent tag on the 3′end) (All from Sigma-Proligo). A 20 µM solution of the DNA was diluted 1:5 in printing buffer (1 M Tris-HCl (pH 7) with 1% v/v N,Ndiisopropylethylamine) and deposited onto the chip surface (either across the whole surface or as a spot deposition over the working electrode area). The chips were then incubated overnight at 37 °C in a dark, humid chamber. Unreacted cross-linker moieties were capped by immersion of the chips in 50 mM 6-amino-1hexanol and 150 mM N,N-diisopropylethylamine in DMF for 2 h, followed by washing with DMF, MeOH, and DI water. Surface coverage of the modified chip surface with the fluorescent DNA was assessed using a Ziess Axiscope fluorescent microscope. Surface modification of both microsquare electrode arrays and microband electrode arrays was undertaken for comparative study, as they both have identical surface chemical properties and the

only difference was in their electrochemical signals (due to differing diffusion profiles, which are dependent on the electrode shape and size). This parallel work placed less restriction on available surfaces for modification studies. Antibody Immobilization onto the Chip Device. The capture antibody (anti-aflatoxin M1) was diluted (96 µg mL-1) with carbonate buffer (0.1 M, pH 9.6), of which 1 µL of the antibody solution was placed onto the device. These were stored overnight at 4 °C in humid conditions to allow covalent attachment via the PDITC cross-linker. The devices were washed twice with 10 mM PBS-T pH 7.4 buffer, once with water using a dispensing bottle, and then shaken dry. After the devices were dried, 3 µL of 0.1% NH4OH in water was added for 60 min at room temperature to deactivate any unreacted PDITC cross-linker and then washed and dried. A volume of 1 µL of 40 µg mL-1 anti-aflatoxin M1 antibody was placed onto the devices and incubated at 37 °C for 2 h in humid conditions. The electrode arrays were then washed and dried as reported above and stored at 4 °C until used. Assay Development for the Chip Device. For the optimization of TMB electrochemical detection using the MEA, differential pulse voltammetry was employed. The working MEA with the immobilized PDITC cross-linker was first capped using 1% NH4OH at room temperature for 1 h before 0.5 mM TMB in 10 mM citrate buffer and 0.1 M KCl was placed onto the electrode surface. The electrode array was connected to an AUTOLAB potentiostat (Eco chemie, The Netherlands) via a custom-made connector and the potential was scanned from -0.5 to 0.5 V using the on-chip pseudoreference and counter electrodes. The immunoreaction of aflatoxin M1 to the activated electrode surface was achieved by placing 1 µL of sample or standard, mixed 1:1 with aflatoxin M1-HRP (diluted 1:10 with 10 mM PBS, pH 7.4) onto the antibody immobilized MEA and incubated at 37 °C for 120 min. The devices were washed twice with 10 mM PBS-Tween (0.05% Tween 20) pH 7.4 buffer and once with water using a dispensing bottle and then was shaken dry. The bound HRP-conjugate was then determined using a TMB/H2O2 solution. This solution was prepared by dissolving 1 mg of TMB in 150 µL of DI water, and 20 µL of this stock solution was mixed with 2 µL of 30% hydrogen peroxide and made up to 1 mL using 10 mM citrate buffer (pH 5.2) containing 0.1 M KCl at 37 °C. A 4 µL aliquot of the TMB/ H2O2 solution was placed onto the MEA immediately prior to analysis. The stock solution of TMB was prepared daily and stored in the dark prior to use. The electrochemical measurements were performed by connecting the microarray to the AUTOLAB potentiostat. A conditioning prepotential was applied first for 5 s at a potential of +268 mV and then the potential was set to +168 mV for measurement (5 min). Preconditioning the electrode as reported above before data collection has been shown previously to increase the signal achieved from the immunoassay.34,42,43 Samples of full fat milk were pretreated by centrifugation at 9000 rpm (5 min), and an aliquot was taken from below the upper fat layer and used in the analysis. Curve fitting of data reported in this paper was carried out using Graphpad Prism version 5.02 from Graphpad software. Surface Analysis of the Microelectrode Array. The surface of the MEAs was characterized by atomic force microscopy (AFM) Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

5293

Figure 1. (a) The three-electrode chips were fabricated with one working electrode area (35 electrodes in the array), a counter electrode, and a reference electrode area. (b) The whole working microelectrode of the untreated surface at 80× magnification using a sFEG. Atomic force microscopy image of a single element of the array for the untreated working microelectrode (image 40 µm × 40 µm).

and scanning electron microscopy (SEM) to monitor the immobilization of the antibody. The SEM images were taken using a Philips XL30 scanning field emission gun (SFEG, U.K.). The AFM images were obtained in a wet environment using a Dimension 3000, from Digital Instruments (now Veeco Instruments, U.K.). The AFM tips used were silicon probes used in the tapping mode. The probes were 225 µm × 38 µm × 7 µm with a typical resonant frequency of 160 kHz. The scan speed applied was between 0.5 and 1 Hz. The surfaces of two MEA devices were analyzed in detail using AFM to monitor immobilization of the antibody. One of these sensors’ surface was prepared by immobilizing the capture antibody prior to the surface analysis. A volume of 1 µL of 96 µg mL-1 of capture antibody (Pierce, U.K.) was placed onto the MEA surface at pH 9.6 and incubated at 4 °C overnight. The surface was washed with 10 mM PBS-T and H2O, and then the excess linker compound was deactivated using 4 µL of 0.1% NH4OH for 1 h at 37 °C. Safety Awareness. All laboratory glassware and consumables which had been contaminated with aflatoxin M1 were stored overnight in 5% sodium hypochlorite, and then acetone was added to make the solution 5% acetone by volume. The decontamination solution was then stored at room temperature for a minimum of 30 min before disposal. RESULTS AND DISCUSSION Cell-on-Chip Device Characterization. Each gold cell-on-achip device consisted of gold counter and reference electrodes with a gold MEA working electrode (Figure 1). Each array consisted of 35 microsquare electrodes with 20 µm × 20 µm dimensions and edge-to-edge spacing of 200 µm (an electrode width to spacing ratio of 10). These dimensions and spacings were chosen to avoid overlapping diffusion layers between neighboring electrodes in the array. The surface of the sensor was first characterized using SEM imaging. These were taken using sFEG rather than a conventional SEM since the sFEG gives better resolution. With the use of a high-resolution scanning electron microscope (sFEG) at a low magnification, images (×80 magnification) of the working MEA 5294

Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

Figure 2. Characterization of Au cell-on-a-chip microelectrodes with 1 mM FcCOOH in 0.01 M PBS using cyclic voltammetry at 5 mV s-1. The influence of different on-chip reference electrode preparations with working electrodes consisting of 35 Au microsquares with 20 µm × 20 µm dimensions and edge-to-edge spacing of 200 µm.

for the microsensors were taken. Figure 1b, clearly shows the layout of the working MEA with the 35 elements etched into the surface and also the image of the single recessed electrode, obtained using AFM. This design was used for the development of the immunosensor. Electrochemical characterization of the cell-on-a-chip microelectrodes was conducted by CV using 1 mM FcCOOH in PBS at 5 mV s-1. Characterization of the working electrode with external counter and reference electrodes was undertaken to ensure that the electrodes were functioning correctly. Further characterization of the cell-on-a-chip devices was undertaken following modification of the on-chip Au reference electrode with Ag and with Ag/AgCl (Figure 2). An apparent shift in the redox potential of the electroactive species can be seen on changing the reference electrode from the pseudoreference unmodified gold to the Ag and Ag/AgCl reference electrodes. The latter modification results in a similar redox potential to the external Ag|AgCl reference electrode. All CVs show characteristic steady-state cyclic voltammograms as expected for microelectrodes with sufficient interelectrode spacing to achieve indepen-

Figure 3. Fluorescence microcope images of surface modified chips using silanization/PDITC chemistry with fluorescently tagged DNA: (a) gold microband electrodes and (b) platinum microsquare electrodes showing the microsquare electrode opening in the silicon nitride layer, underlying Pt electrode, and edge of spot deposition.

dent diffusion profiles for each element of the array and gave currents in agreement with established models for diffusioncontrolled currents at microelectrodes.8-10 Since the choice of the reference or pseudoreference electrode did not influence the magnitude of the current, only its position on the potential axis, the use of a gold pseudoreference electrode for subsequent measurements was chosen. This simplified the electrochemical device preparation by avoiding the necessity to prepare the onchip Ag and Ag/AgCl electrodes. Device Surface Modification. Where small arrays of microelectrodes are used, the total electrode surface area is relatively small, and thus for biosensor applications, a limited amount of biorecognition material can be immobilized. This further limits the information generating capacity of the sensor and compromises the sensitivity. In order to improve the biocapture capacity and sensitivity of the chips, modification of the sensor insulating area (silicon nitride) around the microelectrodes was investigated. Aside from the larger immobilization area, a further advantage of modification to this area is that the electrode surface remains less affected by the surface modifications, thus the signal is less attenuated by the immobilized reagents. The modification of the chips using the amino-silane anchor (APTES) and cross-linker (PDITC) was carried out on the sensors surfaces. The chemistries were applied to the whole chip surface and the surface coverage assessed using fluorescently tagged DNA. These studies were also conducted with gold microband electrodes consisting of four 50 µm × 500 µm bands with 500 µm edge-to-edge separation and platinum microsquare electrodes for comparison of sensor performance. The modification provided a more hydrophobic surface, hence spot depositions could be more precisely and reproducibly applied to the chip. Figure 3 shows the fluorescent microscopic images, demonstrating successful immobilization of the DNA onto the silicon nitride surface. Immobilization was investigated by coverage of the entire chip surface, i.e., the whole of the three-electrode cell-on-chip surface (in this case examined using Au microband electrodes, Figure 3a), and by spot deposition of the DNA over the working electrode area only (tested using the Pt microsquare electrodes, Figure 3b). The edge of the spot can be clearly seen in Figure 3b, indicating successful modification of the surface in that area of the chip. However, it is less clear from these fluorescence images if the metal electrode surfaces were also modified.

The aim of the surface modification was for attachment of the biorecognition material mainly on the surrounding silicon nitride layer; therefore, an assessment of the surface coverage on the gold and platinum electrode surfaces was carried out by voltammetry. The most widely used method to assess the extent of electrode surface coverage by DNA is based on cyclic voltammetry of ferricyanide.31 Shielding of the ferro-/ferricyanide ions by the immobilized DNA can be attributed to a combination of physical coverage of the electrode by the DNA and electrostatic repulsion between the negatively charged redox couple ions and the DNA phosphate backbone.32 Figure 4 shows the cyclic voltammetry of ferricyanide using the DNA-modified gold and platinum MEAs. For each chip, the sensor response was assessed using the onchip reference and counter electrodes and also using external reference and counter electrodes. The response of the DNAmodified platinum electrodes shows a reduced current with respect to the unmodified platinum, suggesting some modification of the electrode surface as well as the surrounding silicon nitride layer around them (Figure 4b). The modified gold electrodes, in contrast, show that the current has not been decreased with respect to the unmodified electrodes, suggesting that the modification procedure has not significantly covered the gold surface (Figure 4a). This is attributed to the reaction of the silane reagent APTES with residual oxides on the surface of the platinum leading to DNA immobilization on the platinum. Gold cell-on-chip sensors were used for subsequent immunosensor development including the use of the on-chip gold pseudoreference electrode. Immunoassay Development. For the detection of aflatoxin M1, a monoclonal antirat antibody was employed as the sensing molecule for the microarray electrode. In order to ensure optimal orientation of the monoclonal antibody on the sensor surface, a polyclonal “capture” antibody was covalently immobilized to the microarray through the PDITC cross-linker chemistry, since covalent immobilizations can cause the antibody to lay in a “side on” orientation33 and therefore results in poor binding efficiency to the analyte. By incorporation of a “capture” antibody, the sensing antibody becomes highly efficient. With the monoclonal sensing antibody immobilized to the surface of the microarray, the detection of aflatoxin M1 was carried out by a competitive reaction between the free aflatoxin M1 in the sample and an aflatoxin M1-horseradish peroxidase conjugate. This assay procedure was developed and Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

5295

Figure 4. Electrochemical characterization of DNA-modified and unmodified chips with 5 mM FeCN in 0.1 M KCl: (a) gold microband electrodes and (b) Pt microsquare electrodes. Data are based on cyclic voltammetry of three separately modified chips (a-c) of Au and Pt with comparison between on-chip and external reference electrodes. Scan rate 5 mV s-1.

characterized in our previous work using screen-printed sensors (macrosensors) for aflatoxin M1 analysis34 before it was transferred after minimal modification to the surface of the microelectrode sensor. The developed method for the macrosensors34 utilized passive adsorption onto a carbon electrode surface as an immobilization protocol. The immobilization method for the microsensors was by covalent attachment using the amino-silane anchor and PDITC cross-linker, which was a deviation from the macrosensor method. The validation of successful immobilization of the antibody onto the surface was performed using atomic force microscopy. After the immobilization of the assay reagents on the microelectrode, a 3D image of the surfaces inside a microsquare electrode was then conducted. This was compared to bare array surface as shown in Figure 1. The results from Figure 5 indicate that the immobilization of the antibodies cause a change of the root mean square (rms) roughness from 1.27 to 2.37 nm (increase of 0.84 nm) agreeing with the observation of other researchers35-37 that upon the addition of protein to a sensor surface the roughness increases. Both the surface roughness and topography indicated quantifiable differences, and visual evidence was seen using 3D imaging and phase control. This shows that the antibodies have (35) Parra, A.; Casero, E.; Pariente, F.; Va´zquez, L.; Lorenzo, E. Sens. Actuators, B 2007, 124, 30–37. (36) Tsai, Y. C.; Huang, J. D.; Chui, C. C. Biosens. Bioelectron. 2007, 22, 3051– 3056. (37) Vianello, F.; Zennaro, L.; Rigo, A. Biosens. Bioelectron. 2007, 22, 2694– 2699.

5296

Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

also passively adsorbed to the gold working MEA while being covalently immobilized to the silicon nitride surrounding. It is estimated that these surface-adsorbed antibodies will have a minimal enhanced effect on the final results due to the large surface of the silicon nitride and small surface of the gold MEA. To eliminate this passive adsorption, the gold microarray surface may need to be blocked using short chain alkane thiols. However, this is not necessary in this work as the effect is small. With the use of the immobilization protocol, several devices were prepared for aflatoxin M1 measurements in buffer as well as spiked commercial milk. Before electrochemical quantification of aflatoxin M1 could be demonstrated, the electrochemical test parameters were optimized for the electrochemical MEA devices. The electrochemical detection of TMB is achieved using either the reduction peak or oxidation peak at +100 mV vs Ag/AgCl38-40 and -100 mV vs Ag/AgCl,41 respectively. Differential pulsed voltammetry was employed to find the maximum detection potential of TMB using the gold microelectrode array. The voltammogram (data not shown) showed that the maximum peak signal occurred at a potential of +168 mV (vs the gold pseudoreference electrode); therefore, the detection of TMB using chronoamperometry was set at +168 mV (vs the gold pseudoreference electrode) and a preconditioning potential of 100 mV above the detection potential (+268 mV vs the gold pseudoreference electrode) was applied for 5 s before chronoamperometric measurement.42,43 The coefficient of variation (%CV) value for current obtained using modified chips (with the amino-silane anchor (APTES) and cross-linker (PDITC)) from TMB detection using chronoamperometry analysis with three replicate electrodes ) 4%. With the use of these settings, a calibration graph for aflatoxin M1 was determined in buffer as well as in spiked milk samples using the microarray immunosensor as shown in Figure 6. Error bars indicate the standard deviation (n ) 3). The results (Figure 6a) show that the microelectrodes are very sensitive and are able to detect levels of toxins lower than the current legislative requirements of 50 ng L-1.25 The analytical sensitivity in buffer is less than 1 ng L-1 (calculated as the amount of aflatoxin M1 needed to produce a 25% decrease in the signal).47 The r2 value for the curve was 0.986 using the Graphpad Prism “one site - fit Ki” curve fitting function. With the devices performing well in pure buffer solutions, further examination was carried out to assess the performance in a milk matrix. With the use of the pretreatment for the milk samples described in the Experimental Section, a calibration curve was established using spiked milk samples. Figure 6b shows that the dynamic detection range in milk samples occurred between 10 and 100 ng L-1 and the analytical sensitivity (calculated as the amount of aflatoxin M1 needed to produce a (38) Badea, M.; Micheli, L.; Messia, M. C.; Candigliota, T.; Marconi, E.; Mottram, T.; Velasco-Garcia, M.; Moscone, D.; Palleschi, G. Anal. Chim. Acta, 2004, 520, 141–148. (39) Fanjul-Bolado, P.; Gonza´lez-Garcı´a, M. B.; Costa-Garcı´a, A. Anal. Bioanal. Chem. 2005, 382, 297–302. (40) Butler, D.; Pravda, M.; Guilbault, G. G. Anal. Chim. Acta 2006, 556, 333– 339. (41) Micheli, L.; Grecco, R.; Badea, M.; Moscone, D.; Palleschi, G. Biosens. Bioelectron. 2005, 21, 588–596. (42) Lu, H.; Conneely, G.; Pravda, M.; Guilbault, G. G. Steroids 2006, 71, 760– 767. (43) Conneely, G.; Aherne, M.; Lu, H.; Guilbault, G. G. Anal. Chim. Acta 2007, 583, 153–160.

Figure 5. (a) The surface roughness taken inside of the untreated element, analyzed by atomic force microscopy. (b) The surface roughness inside a treated element, analyzed by atomic force microscopy. For the treated microelectrode, 1 µL of 96 µg mL-1 dilution of capture antibody was immobilized at 4 °C overnight, and then excess linker compound was deactivated using 4 µL of 0.1% NH4OH for 1 h at 37 °C and AFM images were taken immediately for the gold microsquare array.

r2 value for the curve was 0.999 using the graphpad prism “one site - fit Ki” curve fitting function. The different response curve shown for analysis in buffer and then in milk is due to the use of different batches of sensors to conduct the analysis which at the time of the experiments were manually assembled. Improvements in assembly will minimize this variability, although clearly on-chip variation is acceptable, as shown in the response curves produced in milk samples. The %CV for the immunoassay on the MEA (n ) 3) was ∼6%. In our previous work for the development of macrosensors,34 milk was found to interfere with the electrochemical measurement using the carbon screen-printed electrode. It was deduced that whey proteins, with a specific focus on R-lactalbumin, were blocking the carbon electrode surface. The macrosensors were therefore blocked using a PVA layer; additionally milk samples required pretreatment with calcium chloride to reduce whey proteins blocking of the electrode surface. With the current protocol utilizing the MEAs, the sensor chips were not blocked by any blocking agent as with the carbon based screen-printed electrodes, no matrix interference was observed, and no pretreatment with calcium chloride was required. For the gold MEAs, blocking of the surface with polymers is not required due to the lower absorptive properties of the gold compared to carbon and perhaps the lower surface area (due to smoothness) of lithographically produced electrodes over screen-printed technology which produced a comparably rough surface. As the research reported here is an initial investigation into developing new technology in food safety testing, the work is still in progress and gives “proof of principle” of the technology for a new and important application. The idea is for testing performed Figure 6. Standard curve for aflatoxin M1 detection using the developed MEA cell-on-chip device (a) in buffer solution and (b) in milk. A volume of 1 µL of the sample or standard + aflatoxin M1 HRP (diluted 1:10 with PBS) was placed and incubated at 37 °C for 120 min before detection with TMB (0.5 mM) and hydrogen peroxide (1 mM) in citrate phosphate buffer pH 5.2 with 0.1 M KCl at +168 mV. Error bars indicate the standard deviation (n ) 3).

25% decrease in the signal)47 at 8 ng L-1 which surpasses previous reported immunosensors for aflatoxins.38,41,44-48 The

(44) Carlson, M. A.; Bargeron, C. B.; Benson, R. C.; Fraser, A. B.; Philips, T. E.; Velky, J. T.; Groopman, J. D.; Strickland, P. T.; Ko, H. W. Biosens. Bioelectron. 2000, 14, 841–848. (45) Nasir, M. S.; Jolley, M. E. J. Agric. Food Chem. 2002, 50, 3116–3121. (46) Gaag, B.; Spath, S.; Dietrich, H.; Stigter, E.; Boonzaaijer, G.; Osenbruggen, T.; Koopal, K. Food Control 2003, 14, 251–254. (47) Ammida, N. H. S.; Micheli, L.; Palleschi, G. Anal. Chim. Acta 2004, 520, 159–164. (48) Chiavaro, E.; Caccgiolo, C.; Berni, E.; Spotto, E. Food Addit. Contam. 2005, 22, 1154–1161.

Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

5297

at the “farm” and as such only full fat commercial fresh milk was tested in this investigation before farm testing can be performed. Future work will investigate analysis of a range of milk types and products. Further sensor optimization is required before a final device can be considered for commercialization, such as the reduction of incubation time of the assay, and stabilization of the immunoreagents on the sensor surface, such as drying the immuno-components with an osmolyte e.g., trehalose or betaine. Few comparable publications for MEA-based immunosensors are available. The most directly comparable report in terms of method and application is that reported by Dill49 who has produced an immunosensor for R1 acid glycoprotein with a detection limit of 5 ng L-1. The device reported here has a detection capability of 8 ng L-1 and operates in competitive immunoassay format. With review of other electrochemical immunosensors for the detection of aflatoxin M1, this surpasses the work of Micheli et al., who achieved limits of detection at 25 ng L-1 using screen-printed technology.41 CONCLUSIONS Gold MEAs (35 elements) with on-chip reference and counter electrodes, fabricated using photolithographic techniques, were used in this investigation to develop an immunosensor for aflatoxin M1 in milk samples. The microarray was characterized using cyclic voltammetry and showed good performance similar to that reported in the literature for MEA devices. Surface (49) Dill, K.; Montgomery, D. D.; Ghindilis, A. L.; Schwarzkopf, K. R.; Ragsdale, S. R.; Oleinikov, A. V. Biosens. Bioelectron. 2004, 20, 736–742.

5298

Analytical Chemistry, Vol. 81, No. 13, July 1, 2009

modification was successfully conducted for the covalent immobilization of the capture antibody. A competitive immunoassay was then performed on the surface of the sensor using the TMB/H2O2 electrochemical detection system with HRP as the enzyme label. The sensors were reproducible and very sensitive giving a detection limit of 8 ng L-1 in milk matrix. The sensor has been designed for one use (disposable); however, our current unpublished work indicates that the chip can be recycled after plasma cleaning. Further optimisation of the assay protocol can result in a lower detection limit. The improvements seen by employing gold microelectrodes rather than screen-printed electrodes for aflatoxin M1 detection indicate that for high sensitivity applications, such as low detection limits, a boundary has been surpassed allowing for development of many new applications which have previously only been achieved using elaborate instrumentation. ACKNOWLEDGMENT The authors thank the European Commission for supporting this work (Project FP6- IST1-508774-IP “GOODFOOD: Food safety and quality with microsystems technology”). Damien W. M. Arrigan thanks the Science Foundation Ireland for ongoing support (Grants 02/IN.1/B84 and 07/IN.1/B967).

Received for review March 10, 2009. Accepted May 11, 2009. AC900511E